Throwing velocity is one of the most important factors for scoring goals in handball. This study aimed to identify the type of throw and procedure for selecting the final test outcome that provide throwing velocity with the greatest reliability. Fifteen experienced handball players and 33 non-experienced participants were tested in two sessions. Each session consisted of 4 trials of 3 different throwing tests (unspecific, 7-meters, and 3-steps). The maximum value of 4 trials, average value of 4 trials, and average value of the 3 best trials were considered. Throwing velocity was highly reliable (coefficient of variation [CV]≤3.3%, intraclass correlation coefficient≥0.89) with the exception of the unspecific throw for the non-experienced group (CV≥5.9%, intraclass correlation coefficient≤0.56). The 3-steps throw (CV=1.7%) was more reliable than the 7-meters throw (CV=2.1%) (CVratio=1.19) and unspecific throw (CV = 3.8%) (CVratio=2.18), the 3 procedures provided a comparable reliability (CV range=2.4−2.6%; CVratio≤1.07), and the experienced group (CV=1.0%) presented a higher reliability than the non-experienced group (CV=4.0%) (CVratio=3.83). These results support the 3-steps throw to maximise the reliability of throwing velocity performance.